المؤلفون: Zhang, Hailin, Liu, Zirui, Chen, Boxuan, Zhao, Yikai, Zhao, Tong, Yang, Tong, Cui, Bin

مصطلحات موضوعية: Computer Science - Machine Learning

الوصف: Recently, the growing memory demands of embedding tables in Deep Learning Recommendation Models (DLRMs) pose great challenges for model training and deployment. Existing embedding compression solutions cannot simultaneously meet three key design requirements: memory efficiency, low latency, and adaptability to dynamic data distribution. This paper presents CAFE, a Compact, Adaptive, and Fast Embedding compression framework that addresses the above requirements. The design philosophy of CAFE is to dynamically allocate more memory resources to important features (called hot features), and allocate less memory to unimportant ones. In CAFE, we propose a fast and lightweight sketch data structure, named HotSketch, to capture feature importance and report hot features in real time. For each reported hot feature, we assign it a unique embedding. For the non-hot features, we allow multiple features to share one embedding by using hash embedding technique. Guided by our design philosophy, we further propose a multi-level hash embedding framework to optimize the embedding tables of non-hot features. We theoretically analyze the accuracy of HotSketch, and analyze the model convergence against deviation. Extensive experiments show that CAFE significantly outperforms existing embedding compression methods, yielding 3.92% and 3.68% superior testing AUC on Criteo Kaggle dataset and CriteoTB dataset at a compression ratio of 10000x. The source codes of CAFE are available at GitHub.

الوصول الحر: http://arxiv.org/abs/2312.03256Test

المؤلفون: Chen, Boxuan, Zhang, Ruifeng, Wang, Shaochu, Zhang, Liqiang, Liu, Yu

مصطلحات موضوعية: Computer Science - Robotics

الوصف: Pedestrian dead reckoning is a challenging task due to the low-cost inertial sensor error accumulation. Recent research has shown that deep learning methods can achieve impressive performance in handling this issue. In this letter, we propose inertial odometry using a deep learning-based velocity estimation method. The deep neural network based on Res2Net modules and two convolutional block attention modules is leveraged to restore the potential connection between the horizontal velocity vector and raw inertial data from a smartphone. Our network is trained using only fifty percent of the public inertial odometry dataset (RoNIN) data. Then, it is validated on the RoNIN testing dataset and another public inertial odometry dataset (OXIOD). Compared with the traditional step-length and heading system-based algorithm, our approach decreases the absolute translation error (ATE) by 76%-86%. In addition, compared with the state-of-the-art deep learning method (RoNIN), our method improves its ATE by 6%-31.4%.

الوصول الحر: http://arxiv.org/abs/2205.10031Test

المؤلفون: Chen, Boxuan, Qin, Shaozheng

المساهمون: National Natural Science Foundation of China, Fundamental Research Funds for Central Universities of the Central South University

المصدر: Psychiatry and Clinical Neurosciences ; volume 78, issue 5, page 271-272 ; ISSN 1323-1316 1440-1819

الإتاحة: https://doi.org/10.1111/pcn.13662Test

المؤلفون: Zhang, Hailin, Liu, Zirui, Chen, Boxuan, Zhao, Yikai, Zhao, Tong, Yang, Tong, Cui, Bin

المصدر: Proceedings of the ACM on Management of Data ; volume 2, issue 1, page 1-28 ; ISSN 2836-6573

الوصف: Recently, the growing memory demands of embedding tables in Deep Learning Recommendation Models (DLRMs) pose great challenges for model training and deployment. Existing embedding compression solutions cannot simultaneously meet three key design requirements: memory efficiency, low latency, and adaptability to dynamic data distribution. This paper presents CAFE, a Compact, Adaptive, and Fast Embedding compression framework that addresses the above requirements. The design philosophy of CAFE is to dynamically allocate more memory resources to important features (called hot features), and allocate less memory to unimportant ones. In CAFE, we propose a fast and lightweight sketch data structure, named HotSketch, to capture feature importance and report hot features in real time. For each reported hot feature, we assign it a unique embedding. For the non-hot features, we allow multiple features to share one embedding by using hash embedding technique. Guided by our design philosophy, we further propose a multi-level hash embedding framework to optimize the embedding tables of non-hot features. We theoretically analyze the accuracy of HotSketch, and analyze the model convergence against deviation. Extensive experiments show that CAFE significantly outperforms existing embedding compression methods, yielding 3.92% and 3.68% superior testing AUC on Criteo Kaggle dataset and CriteoTB dataset at a compression ratio of 10000x. The source codes of CAFE are available at GitHub.

الإتاحة: https://doi.org/10.1145/3639306Test

المؤلفون: Potort, Francesco, Torres-Sospedra, Joaquín, Quezada Gaibor, Darwin, Jiménez, Antonio Ramón, Seco, Fernando, Perez-Navarro, Antoni, Ortiz, Miguel, Zhu, Ni, Renaudin, Valerie, Ichikari, Ryosuke, Shimomura, Ryo, Kaichi, Tomoya, Zhou, Baoding, Liu, Xu, Gu, Zhining, Yang, Chengjing, Wu, Zhiqian, Xie, Doudou, Huang, Can, Zheng, Lingxiang, Peng, Ao, Jin, Ge, Wang, Qu, Luo, Haiyong, Xiong, Hao, Bao, Linfeng, Zhang, Pushuo, Zhao, Fang, Yu, Chia-An, Hung, Chung-Hao, Antsfeld, Leonid, Chidlovskii, Boris, Jiang, Haitao, Xia, Ming, Yan, Dayu, Li, Yuhang, Dong, Yitong, Silva, Ivo, Pendão, Cristiano, Meneses, Filipe, Nicolau, Maria João, Costa, António, Moreira, Adriano, De Cock, Cedric, Plets, David, Opiela, Miroslav, Dzama, Jakub, Zhang, Liqiang, Li, Hu, Chen, Boxuan, Liu, Yu, Yean, Seanglidet, Lim, Bo Zhi, Teo, Wei Jie, Lee, Bu Sung, OH, HL, ohta, nozomu, Nagae, Satsuki, Kurata, Takeshi, dongyan, wei, Ji, Xinchun, Zhang, Wenchao, Kram, Sebastian, Stahlke, Maximilian, Mutschler, Christopher, Crivello, Antonino, Barsocchi, Paolo, GIROLAMI, MICHELE, Palumbo, Filippo, Chen, Ruizhi, Wu, Yuan, Li, Wei, Yu, Yue, Xu, Shihao, Huang, Lixiong, Liu, Tao, Kuang, Jian, Niu, Xiaoji, Yoshida, Takuto, Nagata, Yoshiteru, Fukushima, Yuto, Fukatani, Nobuya, Hayashida, Nozomi, Asai, Yusuke, Urano, Kenta, Ge, Wenfei, Lee, Nien-Ting, Fang, Shih-Hau, Jie, You-Cheng, Young, Shawn-Rong, Chien, Ying-Ren, Yu, Chih-Chieh, Ma, Chengqi, Wu, Bang, Zhang, Wei, Wang, Yankun, Fan, Yonglei, Poslad, Stefan, Selviah, David, Wang, Weixi, Yuan, Hong, Yonamoto, Yoshitomo, Yamaguchi, Masahiro

المساهمون: Universitat Oberta de Catalunya (UOC)

مصطلحات موضوعية: indoor positioning and navigation, evaluation, smartphone-based positioning, foot-mounted IMU, positioning in industrial scenarios and factories, vehicle-positioning, navegación y posicionamiento en interiores, evaluación, posiciónamiento basado en teléfonos inteligentes, IMU con patas, posicionamiento en escenarios industriales y fábricas, posicionamiento del vehículo, navegació i posicionament en interiors, evaluació, posicionament basat en teléfons inteligents, IMU amb cames, posicionament en escenaris industrials i fabriques, posicionament del vehicle, electronics in navigation, electrònica en la navegació, electrónica en la nevagación

الوصف: Every year, for ten years now, the IPIN competition has aimed at evaluating real-world indoor localisation systems by testing them in a realistic environment, with realistic movement, using the EvAAL framework. The competition provided a unique overview of the state-of-the-art of systems, technologies, and methods for indoorpositioning andnavigationpurposes.Throughfaircomparison of the performance achieved by each system, the competition was able to identify the most promising approaches and to pinpoint the most critical working conditions. In 2020, the competition included 5 diverse off-site off-site Tracks, each resembling real use cases and challenges for indoor positioning. The results in terms of participation and accuracy of the proposed systems have been encouraging. The best performing competitors obtained a third quartile of error of 1m for the Smartphone Track and 0.5m for the Footmounted IMU Track. While not running on physical systems, but only as algorithms, these results represent impressive achievements

وصف الملف: application/pdf

العلاقة: IEEE Sensors Journal, 2022, 22(6); https://www.doi.org/10.1109/JSEN.2021.3083149Test; http://hdl.handle.net/10609/147005Test

الإتاحة: https://doi.org/10.1109/JSEN.2021.3083149Test
http://hdl.handle.net/10609/147005Test

المؤلفون: Potorti, Francesco, Torres-Sospedra, Joaquín, Quezada-Gaibor, Darwin, Jimenez, Antonio Ramon, Seco, Fernando, Perez-Navarro, Antoni, Ortiz, Miguel, Zhu, Ni, Renaudin, Valerie, Ichikari, Ryosuke, Shimomura, Ryo, Ohta, Nozomu, Nagae, Satsuki, Kurata, Takeshi, Wei, Dongyan, Ji, Xinchun, Zhang, Wenchao, Kram, Sebastian, Stahlke, Maximilian, Mutschler, Christopher, Crivello, Antonino, Barsocchi, Paolo, Girolami, Michele, Palumbo, Filippo, Chen, Ruizhi, Wu, Yuan, Li, Wei, Yu, Yue, Xu, Shihao, Huang, Lixiong, Liu, Tao, Kuang, Jian, Niu, Xiaoji, Yoshida, Takuto, Nagata, Yoshiteru, Fukushima, Yuto, Fukatani, Nobuya, Hayashida, Nozomi, Asai, Yusuke, Urano, Kenta, Ge, Wenfei, Lee, Nien-Ting, Fang, Shih-Hau, Jie, You-Cheng, Young, Shawn-Rong, Chien, Ying-Ren, Yu, Chih-Chieh, Ma, Chengqi, Wu, Bang, Zhang, Wei, Wang, Yankun, Fan, Yonglei, Poslad, Stefan, Selviah, David R., Wang, Weixi, Yuan, Hong, Yonamoto, Yoshitomo, Yamaguchi, Masahiro, Kaichi, Tomoya, Zhou, Baoding, Liu, Xu, Gu, Zhining, Yang, Chengjing, Wu, Zhiqian, Xie, Doudou, Huang, Can, Zheng, Lingxiang, Peng, Ao, Jin, Ge, Wang, Qu, Luo, Haiyong, Xiong, Hao, Bao, Linfeng, Zhang, Pushuo, Zhao, Fang, Yu, Chia-An, Hung, Chun-Hao, Antsfeld, Leonid, Silva, Ivo Miguel Menezes, Pendão, Cristiano Gonçalves, Meneses, Filipe, Nicolau, Maria João, Costa, António, Moreira, Adriano, Cock, Cedric De, Plets, David, Opiela, Miroslav, Jakub Džama, Zhang, Liqiang, Li, Hu, Chen, Boxuan, Liu, Yu, Yean, Seanglidet, Lim, Bo Zhi, Teo, Wei Jie, Lee, Bu Sung, Oh, Hong Lye

مصطلحات موضوعية: Sensor phenomena and characterization, Indoor navigation, Testing, Standards, Satellite broadcasting, Recurrent neural networks, Received signal strength indicator, Indoor positioning and navigation, evaluation, smartphone-based positioning, foot-mounted IMU, positioning in industrial scenarios and factories, vehicle-positioning, Ciências Naturais::Ciências da Computação e da Informação, Science & Technology

الوصف: Every year, for ten years now, the IPIN competition has aimed at evaluating real-world indoor localisation systems by testing them in a realistic environment, with realistic movement, using the EvAAL framework. The competition provided a unique overview of the state-of-the-art of systems, technologies, and methods for indoor positioning and navigation purposes. Through fair comparison of the performance achieved by each system, the competition was able to identify the most promising approaches and to pinpoint the most critical working conditions. In 2020, the competition included 5 diverse off-site off-site Tracks, each resembling real use cases and challenges for indoor positioning. The results in terms of participation and accuracy of the proposed systems have been encouraging. The best performing competitors obtained a third quartile of error of 1 m for the Smartphone Track and 0.5 m for the Foot-mounted IMU Track. While not running on physical systems, but only as algorithms, these results represent impressive achievements. ; Track 3 organizers were supported by the European Union’s Horizon 2020 Research and Innovation programme under the Marie Skłodowska Curie Grant 813278 (A-WEAR: A network for dynamic WEarable Applications with pRivacy constraints), MICROCEBUS (MICINN, ref. RTI2018-095168-B-C55, MCIU/AEI/FEDER UE), INSIGNIA (MICINN ref. PTQ2018-009981), and REPNIN+ (MICINN, ref. TEC2017-90808-REDT). We would like to thanks the UJI’s Library managers and employees for their support while collecting the required datasets for Track 3. Track 5 organizers were supported by JST-OPERA Program, Japan, under Grant JPMJOP1612. Track 7 organizers were supported by the Bavarian Ministry for Economic Affairs, Infrastructure, Transport and Technology through the Center for Analytics-Data-Applications (ADA-Center) within the framework of “BAYERN DIGITAL II. ” Team UMinho (Track 3) was supported by FCT—Fundação para a Ciência e Tecnologia within the R&D Units Project Scope under Grant UIDB/00319/2020, and the Ph.D. ...

وصف الملف: application/pdf

العلاقة: info:eu-repo/grantAgreement/EC/H2020/813278/EU; info:eu-repo/grantAgreement/FCT/6817 - DCRRNI ID/UIDB%2F00319%2F2020/PT; info:eu-repo/grantAgreement/FCT/POR_NORTE/PD%2FBD%2F137401%2F2018/PT; https://ieeexplore.ieee.org/document/9439493Test; F. Potortì et al., "Off-Line Evaluation of Indoor Positioning Systems in Different Scenarios: The Experiences From IPIN 2020 Competition," in IEEE Sensors Journal, vol. 22, no. 6, pp. 5011-5054, 15 March15, 2022, doi:10.1109/JSEN.2021.3083149.; https://hdl.handle.net/1822/82092Test

الإتاحة: https://doi.org/10.1109/JSEN.2021.3083149Test
https://hdl.handle.net/1822/82092Test

المؤلفون: Ouyang, Ping, Chen, Boxuan, Shen, Quanxi, Zhu, Wanxu

المساهمون: national natural science foundation of china, Key R&D Projects in the Guangxi Autonomous Region, Major Construction Program of the Science and Technological Innovation Base in the Guangxi Autonomous Region

المصدر: Advances in Mechanical Engineering ; volume 14, issue 10, page 168781322211278 ; ISSN 1687-8132 1687-8140

مصطلحات موضوعية: Mechanical Engineering

الوصف: In this study, a transient heat flow model has been established for parallel wire or strand cables at high temperature by employing the lumped thermal mass approach, and the numerical solution of the surface temperature as a function of time in each layer of the steel wire or strand inside the cable was calculated. Accuracy of the theoretical method is verified through uniform heating test of 73Φ15.7 mm steel strand cable. The results calculated show that temperature field inhomogeneity of cable section is overestimation on the condition that heat conduction inside the cable is not considered. Considering heat conduction or not, the maximum temperature difference of the core steel stand at the same time point is 373 . Unprotected 73Φ15.7 mm cable is damaged in only 12 min in UL1709 fire, and arrangement of fire protection layer around cable can effectively retard the temperature rise of cable surface. The experimental results are in good agreement with the theoretical calculation values in early stage of fire, and the temperature difference between the two is within 10%. Besides, the numerical calculations were analyzed in accordance with the fire protection requirements limiting the surface temperature of cables. It was observed that the minimum thickness of the fire protection layer required to meet the PTI DC45.1-12 standard was linearly related to the numerical value of the section factor of the outermost layer of steel wires or steel strands in the equivalent model, and the slope of the function was approximately equal to the conduction coefficient of fire protection layer, as the cable was subjected to fire for 30 min. Further, based on this, a simple method was proposed to calculate the minimum thickness of the fire protection layer for parallel wire or strand cables.

الإتاحة: https://doi.org/10.1177/16878132221127830Test

المؤلفون: HUANG, ShiZhang, LI, XinZhu, FU, Song, WANG, WeiRong, CHEN, BoXuan, LI, QiBing

المصدر: SCIENTIA SINICA Physica, Mechanica & Astronomica ; volume 50, issue 10, page 104708 ; ISSN 1674-7275

الإتاحة: https://doi.org/10.1360/sspma-2020-0073Test

المؤلفون: Wang, Zhang, Liu, Haiyue, Wang, Fengyan, Yang, Yuqiong, Wang, Xiaojuan, Chen, Boxuan, Stampfli, Martin R., Zhou, Hongwei, Shu, Wensheng, Brightling, Christopher E., Liang, Zhenyu, Chen, Rongchang

المصدر: Frontiers in Microbiology ; volume 11 ; ISSN 1664-302X

مصطلحات موضوعية: Microbiology (medical), Microbiology

الإتاحة: https://doi.org/10.3389/fmicb.2020.01758Test

المؤلفون: Wu, Yue, Liu, Xiuming, Chen, Boxuan, Li, Qingpeng, Luo, Peng, Pronk, ADC Arno

المصدر: ISSN:0926-5805.

الوصف: Ice structures are widely constructed in cold regions as landscapes or shelters. However, owing to the poor mechanical properties of the material and the masonry construction method, the appearance modelling and building scale of ice structures are limited. It is necessary to discover an innovative high-performance ice material and to establish a large-span ice shell construction method. Based on previous studies, a free-form ice composite shell is designed and constructed with an inflatable formwork. A combined optimization algorithm is proposed to determine a reasonable inflatable formwork for the construction, and exhaustive structural investigations on the behaviour of the ice composite shell under multiple load conditions are conducted. Subsequently, some key technologies for the construction process and health monitoring of the ice composite shell are discussed. The results show that the ice composite material is suitable for the construction of large-span ice shells. The morphological design objectives of the free-form ice shell are achieved based on the previously mentioned optimization algorithm, and the structural analysis provides a theoretical foundation for the structural safety and construction quality of the ice shell. Moreover, the effects of the ambient temperature and solar radiation are not negligible, and further research is required.

وصف الملف: application/pdf

العلاقة: http://repository.tue.nl/912342Test

الإتاحة: http://repository.tue.nl/912342Test